Introduction to heatmaps

What are heatmaps?

• ideal for showing 3 variables in one plot. E.g. Calls per hour of day: Call time
• heatmaps many times contain dendrograms. Dendograms are formed by hierachical chlistering. Shows relationships between objects by branches. In heatmaps, dendrograms shows how similar samples are. The color indicates which attributes of the data are responnsible for the clustering of the dendrogram.

First dataset: German weather data from the DWD

• library(tidyverse) is a collection of useful packages, many of which we have used before such as dplyr or ggplot2. Once you load the tidyverse package, you do not need to load the other packages seperately.

• if the tidyverse installation failed, we can also install/load the single libraries dplyr, ggplot2, tibble, pheatmap

• the months of the data are indexed by numbers. E.g. 1 means January, 2 means Feb etc

• for creating the heatmap, we need to convert the data into the wide format (columns are months, rows are the years) using pivot_wider()

• the column_to_rownames() function from the tibble library converts a specified column into the rownames

• colnames() function can be used to rename the column names

## load tidyverse packages
library(tibble)
library(ggplot2)
library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(pheatmap)
library(tidyr)

## load data
meantemp_germany <- read.csv(file = "~/PI/Lectures/R_course/3rd_cycle/02_heatmaps/meantemp_germany.csv", sep = ",")

## modify data (wide format, all data needs to be numeric)

meantemp_germany_clean <- meantemp_germany %>% 
  select(year, month, mean_temp_ger) %>% 
  filter(year >= 1972) %>% 
  pivot_wider(names_from = month, values_from = mean_temp_ger)

meantemp_germany_clean <- column_to_rownames(meantemp_germany_clean, 'year')

colnames(meantemp_germany_clean) <- c("January", "February", "March", "April", "May", "June", "July", "August", "September", "October", "November", "December")

Heatmap generation

• now that the data is in the right shape, a heatmap can be easily

#default pheatmap function
pheatmap(meantemp_germany_clean)

• if you do not want the columns to be clustered, you can turn the dendrogram off for the rows and columns

#turn off default clustering
pheatmap(meantemp_germany_clean, cluster_rows = FALSE, cluster_cols = FALSE)

pheatmap(meantemp_germany_clean, cluster_rows = FALSE, cluster_cols = FALSE, border_color = "white")

• you can also have gaps between the columns and/or rows

#add gaps between columns and rows
pheatmap(meantemp_germany_clean, cluster_rows = FALSE, cluster_cols = FALSE, border_color = "white", gaps_col = c(3, 6, 9), gaps_row = c(9, 19, 29, 39, 49))

• if you want to annotate columns you can generage a new dataframe and use the annotation_col argument

##add annotation
#create annotation data frame

season_df <- data.frame("season" = c("Winter", "Winter", "Spring", "Spring", "Spring", "Summer", "Summer", "Summer", "Autumn", "Autumn", "Autumn", "Winter"))

#change row names 
row.names(season_df) <- c("January", "February", "March", "April", "May", "June", "July", "August", "September", "October", "November", "December")


pheatmap(meantemp_germany_clean, cluster_cols = F, cluster_rows = F, border_color = "white", gaps_col = c(3, 6, 9), gaps_row = c(9, 19, 29, 39, 49), annotation_col = season_df)

• in order to change angle of the x-axis label, use the angle_col argument

# change label angle of columns
pheatmap(meantemp_germany_clean, cluster_cols = F, cluster_rows = F, border_color = "white", gaps_col = c(3, 6, 9), gaps_row = c(9, 19, 29, 39, 49), annotation_col = season_df, angle_col = 45)

• if you want the values of the data to be shown in the heatmap, use the argument display_numbers

# change label angle of columns
pheatmap(meantemp_germany_clean, cluster_cols = F, cluster_rows = F, border_color = "white", gaps_col = c(3, 6, 9), gaps_row = c(9, 19, 29, 39, 49), annotation_col = season_df, angle_col = 45, display_numbers = T)

• in order to save the heatmap (or any plot) - use the ggsave() function from the ggplot2 library. For that, we need to assign the heatmap to a variable so that R knows what you want to save

# assign heatmap to a variable
plot <- pheatmap(meantemp_germany_clean, cluster_cols = F, cluster_rows = F, border_color = "white", gaps_col = c(3, 6, 9), gaps_row = c(9, 19, 29, 39, 49), annotation_col = season_df, angle_col = 45, display_numbers = T)

# save the plot as e.g. png
ggsave(
  "~/Desktop/heatmap.png",
  plot = plot,
  width = 15,
  height = 20,
  units = c("cm"),
)

More information

• the following tricks were not covered in the

## change annotation colors
#create color list
season_color = list(season = c("Winter" = "blue", "Spring" = "green", "Summer" = "red", "Autumn" = "yellow")) 

pheatmap(meantemp_germany_clean, cluster_cols = F, cluster_rows = F, border_color = "white", gaps_col = c(3, 6, 9), gaps_row = c(9, 19, 29, 39, 49), annotation_col = season_df, annotation_colors = season_color) 

#skip row names
pheatmap(meantemp_germany_clean, cluster_cols = F, cluster_rows = F, border_color = "white", gaps_col = c(3, 6, 9), gaps_row = c(9, 19, 29, 39, 49), annotation_col = season_df, angle_col = 45, display_numbers = T, labels_row = c("1972", "", "1974", "", "1976", "", "1978", "", "1980", "", "1982", "", "1984", "", "1986", "", "1988", "", "1990", "", "1992", "", "1994", "", "1996", "", "1998", "", "2000", "", "2002", "", "2004", "", "2006", "", "2008", "", "2010", "", "2012", "", "2014", "", "2016", "", "2018", "", "2020", "", "2022"))

Sequencing data

• Different patients are seperated in different columns
  
• HCC = hepatocellular carcinoma (liver cancer), CRC = colorectal carcinoma
  
• data needs to be cleaned up (as described above) in order to use the object as the input for: get rid of unneccessary columns and rename the rownames

# load data
seq_data_clean <- read.csv(file = "~/PI/Lectures/R_course/3rd_cycle/02_heatmaps/seq_data_clean.csv", sep = ",")

# get rid of unneccessary X column
seq_data_clean$X <- NULL

# use gene names as rownames
seq_data_clean <- column_to_rownames(seq_data_clean, "gene_name")

# generate heatmap
pheatmap(seq_data_clean, scale = "row")